Re: [PATCH v3] net: remove an assert call in eth_get_gso_type

[Jason Wang]

On 2020/10/26 下午5:59, Peter Maydell wrote:
> On Wed, 21 Oct 2020 at 10:23, P J P <ppandit@redhat.com> wrote:
> > +-- On Wed, 21 Oct 2020, Jason Wang wrote --+
> > | It should not be a guest error, since guest is allowed to send a packet
> > | other than IPV4(6).
> >
> > * Ah...sigh! :(
> >
> > * I very hesitantly used guest_error mask, since it was g_assert-ing before.
> >    To me both guest_error and log_unimp seem mismatching. Because no GSO is
> >    also valid IIUC. That's why in patch v2 I used plain qemu_log(). But plain
> >    qemu_log is also not good it seems.
> Well, as I said last time round, the right function depends on what
> is going on here. If this is "the fallback code path is fine, it
> might just be a bit inefficient", then either no logging or use
> a tracepoint. If this is "the guest is allowed to send this packet
> but we're going to mishandle it" then use LOG_UNIMP.


Ok, rethink about this. I think at least 802.1Q is a valid option for GSO.

So I decide to apply the path with LOG_UNIMP.

Thanks


> thanks
> -- PMM